“If anyone is able to show me that what I think or what I do is wrong, I will gladly change, because I am seeking the truth, by which no one has ever been truly hurt.” —Marcus Aurelius, Meditations, Book VI, 21
Everyone wants to use AI. But where, exactly, should do you use it? This is the practical question that CEOs and product managers ask us all the time. They are enthusiastic about applying AI, but also cautious. They want to use it everywhere but know how to prioritize impact over novelty.
One of the main ways Mainsail AI Labs helping our portfolio companies understand and choose which problems to solve with AI is a process called Value Stream Mapping (VSM).
In this article, I’ll review two high-impact examples of how we’ve seen VSM used in practice to make AI decisions:
- Acceleration of engineering throughput by identifying the true stress point.
- Reduce churn by optimizing customer onboarding.
These are two levers that most software companies can leverage to streamline processes with AI, gain efficiencies, and deliver more tangible value to customers.
Seek first to understand, then to improve: What is value stream mapping?
Value stream mapping is a visual, measurable method for gaining visibility into how work is progressing in an organization, where it is slowing or stopping, and where it can be improved.
VSM involves mapping the complete flow of this work, from initial initiation to delivery to the customer, and every step in between. But it’s not just a diagram of boxes and arrows: VSM captures real data:
- Cycle time
- Waiting time
- Speed
- Quality
VSM often reveals that as software companies scale, delays are not due to code issues or individual efforts. Rather, the delays lie in silent handoffs, problem solving without ownership, and misaligned assumptions. VSM codifies each of these elements and allows you to focus on reducing these bottlenecks.
Traditionally, VSM was used in the manufacturing sector. But like many ideas born in the factory, it is even more powerful in software development, where complexity increases exponentially.
VSM can also be applied externally to optimize processes such as customer onboarding, adoption, renewals, go-to-market workflows and retention loops.
Either way, the result is a better idea of what your problems and opportunities really look like before you set out to solve them.
As Epictetus said: “First tell yourself what you would be, then do what you have to do.”
Example #1: AI in engineering
The challenge: Engineering teams often start with AI coding assistants to help their developers work faster. But coding isn’t always the cause of inefficiency, and improving code output without fixing other workflow issues can make the problem worse. For most teams, the real delays come from waiting for code reviews, QA or testing pipelines, or deployment approvals.
Consider this… making developers 50% faster doesn’t make customers 50% happier if the code is waiting in QA. Instead, it increases the burden of quality assurance and reduces the overall capacity of the organization.
And if quality control cuts corners to catch up, you risk delivering buggy products. Faster code just pushes more work downstream to another process that’s already struggling to keep up.
You can only move at the speed corresponding to your slowest step in the chain.
How value stream mapping can help: By mapping your process from idea to production – including key data points such as cycle and wait times – bottlenecks become clear and you can identify areas where AI could have the most impact. In our experience, this is almost always about waits, review queues, QA delays, and deployment hurdles, not coding.
Here is a typical value stream map for engineering:
What it reveals as opportunities for AI:
- Code review bottleneck: If peer reviews take too long, use automated AI-assisted code reviews to complete the first pass immediately after code is checked in.
- Quality assurance bottleneck: If manual QA is the longest process in your entire value chain, leverage AI for automated test generation to cover common and time-consuming test plans.
With these issues resolved, AI coding assistants can finally deliver value through a streamlined process.
Example #2: AI in customer onboarding
Not all AI use cases are in engineering. One of the most overlooked opportunities for AI In software companies, customer onboarding is important, and VSM can help here too.
When we mapped a company’s onboarding flow, we identified significant friction points:
- Delays between signing and kickoff
- Transfer gaps between sales and customer success
- Repetitive requests for the same customer data
- Delays during integration due to manual data imports
Some of them could be solved through workflow optimization. However, several of them had clear paths to AI or automation:
- Intelligent routing of customer requests
- Automated documentation delivery
- Pattern recognition to flag risky accounts before they’re blocked
- Automated data imports leveraging existing APIs
With improved visibility, even basic automation reduced time to value and improved retention for this company. As with engineering, VSM revealed where the problem was and how AI could help alleviate it. Best of all, the team achieved this automation without having to write new code: a win-win situation for the company.
Why VSM is important for your AI strategy
At Mainsail AI Labs, we believe that setting goals is a critical step toward strategically applying AI in any part of your organization. VSM not only helps to accurately identify where to use AI in your process, but, with this knowledge, you can then set measurable goals and identify how AI can help you achieve them.
Example goals for AI in engineering:
- Double the number of pull requests per developer each period.
- Increase code coverage for test cases to 90%.
- Reduce review time by 50% from the time code is completed to the time it is released to production.
- Make sure every feature is documented.
Example goals for AI in customer success:
- Reduce onboarding time by 50%.
- Automatically respond to incoming product help requests 90% of the time.
- Automate material generation for quarterly reviews using customer usage data.
5 steps to start using value stream mapping
You don’t need a team of consultants to do this. Start small:
- Choose a workflow: integration, deployment, or whatever keeps you up at night.
- List each step and who owns it.
- Mark delays, transfers, waits and reworks.
- Estimate the time spent on each part.
- Look for friction, then ask: How can AI help?
Pro tip: Hire the people who do the work. What they say versus what you think might surprise you.
AI ROI starts with clarity
AI is powerful, but not everywhere or at once. If you want AI to accelerate your business, you need to understand your business. It starts by mapping how value flows there and finding the points where it gets stuck.
Once you have this visibility, AI becomes a strategy, not just an experiment.
Marcus Aurelius wrote that he sought the truth, by which no one was ever harmed. Value stream mapping is simply a tool for finding that truth in your engineering, product and other cycles, eliminating the guesswork and assumption that AI is a magic wand.
Need help finding your truth? Leverage Mainsail AI Labs to quickly and efficiently discover the best use cases for AI in your business.